chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

journal homepage: [Link]/locate/cherd

Design and analysis of an ethyl benzene

production process using conventional distillation
columns and dividing-wall column for multiple
objectives

Qi Han Ng, Shivom Sharma, Gade Pandu Rangaiah ∗

Department of Chemical & Biomolecular Engineering, National University of Singapore, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Downstream separation in the ethyl benzene (EB) production is energy intensive due to the
Received 11 May 2016 use of multiple distillation columns. One technology to achieve significant energy and cap-
Received in revised form 17 August ital cost savings for this separation, is the use of dividing-wall column (DWC) to replace
2016 conventional columns. In the present study, a typical industrial EB process including reac-
Accepted 23 October 2016 tion, separation and heat exchangers, is first simulated rigorously and modified to improve
Available online 31 October 2016 the efficiency and performance of both individual units and the overall plant. DWC is inte-
grated into the EB process, and its performance is compared with the base case of utilizing
Keywords: only conventional columns for downstream separation. All simulations are performed using
Ethyl benzene Aspen HYSYS, and design data used is appropriately validated for realistic simulation. Sub-
Distillation sequently, sensitivity analysis is performed on both EB designs for a number of design
Dividing-wall column variables. Objectives selected for sensitivity analysis are net present value, total capital
Net present value investment and benzene loss. Results show that integrating DWC is not only viable but
Total capital investment also offers potential to reduce capital investment, decrease benzene loss and increase profit
compared to a conventional column design for the EB process.
© 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction to extreme operating conditions resulting in catalyst deactivation and

product contamination (Welch et al., 2005). Hence, the latest trend is
Production of ethyl benzene (EB) is vital due to the importance of prod- towards liquid-phase alkylation using zeolite catalysts, which repre-
ucts that can be derived from it. Over 90% of EB is utilized to produce sents around 23% of the market globally and is increasing worldwide
styrene monomer, of which 65% is to produce polystyrene, and the (Netzer and Ghalayini, 2002), offering better catalyst life and optimal
remaining is to produce other polymers/copolymers (Gerzeliev et al., thermal control (Perego and Ingallina, 2002). Commercial technologies
2011). These polymers/copolymers have a large range of applications of liquid-phase zeolite-based alkylation processes include the Lum-
(Collins and Richey, 1992). Historically, there have been both catalyst mus/UOP EBOneTM and the Mobil/Badger EBMaxTM , which allow for
upgrades and process modifications for EB production. The previous high catalyst stability and relatively low benzene recycle rates. Addi-
use of mineral acids such as aluminium chloride is now replaced by tional details and benefits of these processes are elaborated in Woodle
zeolite-based catalysts, which offer improved EB selectivity (Ebrahimi (2006) and Welch et al. (2005). In 1994, CDTech introduced a mixed
et al., 2011), a more environmentally friendly option (Odedairo and Al- liquid-vapour phase alkylation process with zeolite catalyst, utilizing
Khattaf, 2010) and reduced equipment corrosion (Gerzeliev et al., 2011). reactive distillation (RD). Although it represents only 2% of the market
Both vapour and liquid-phase alkylation have used zeolite catalysts. globally, it is rapidly increasing (Netzer and Ghalayini, 2002). Further, Qi
Although vapour-phase alkylation represents nearly 50% of EB mar- and Zhang (2004) pointed that RD is still in the developmental phase for
ket globally, it is declining (Netzer and Ghalayini, 2002), primarily due EB production. Dow Chemical and Indian Petrochemicals Corporation

∗
Corresponding author.
E-mail addresses: gprangaiah@[Link], GPRangaiah@[Link], chegpr@[Link] (G.P. Rangaiah).
[Link]
0263-8762/© 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 143

to unavailability of reaction kinetics), thus requiring fewer distillation

columns for separation. Ebrahimi et al. (2011, 2012) suggested a dif-
Nomenclature
ferent design for EB process with three packed-bed adiabatic reactors
Acronyms in the reaction section and four columns in the separation section. In
B Benzene all designs, alkylation and trans-alkylation reactions occur in separate
CDC Conventional distillation column reactors to improve process efficiency (Gerzeliev et al., 2011). Guidelines
CEPCI Chemical engineering plant cost index and ranges for operating variables such as benzene to ethylene (B/E)
ratio and temperature are detailed in Woodle (2006) and in Al-Kinany
CSTR Continuous stirred tank reactor
et al. (2012) with reasons elaborated.
DEB Di-ethyl benzene
Various reaction kinetic models have been proposed for liquid-
DWC Dividing-wall column
phase alkylation and/or trans-alkylation reactions by Ganji et al. (2004),
DV Decision variable
Luyben (2002, 2010) and You et al. (2006). Ganji et al. (2004) have also
E Ethylene provided plant data of such liquid-phase reactions, which are useful for
EB Ethyl benzene validation and selection of a kinetic model for reactor design. EB pro-
PEB Poly-ethyl benzene cess with conventional distillation columns (CDCs) was studied earlier.
PF Pre-fractionator Luyben (2010) considered total annual cost as an objective in EB plant
PFR Plug flow reactor design. Ebrahimi et al. (2011) maximized EB selectivity by varying both
PR Peng–Robinson ethylene feed flow rate and reactor temperature. In the later work of
RD Reactive distillation Ebrahimi et al. (2012), objective was the concentration of trans-DEB,
which is related to EB selectivity. However, the objectives employed in
SRK Soave–Redlich–Kwong
all these studies do not reflect the performance of the entire process.
VLE Vapour–liquid equilibrium
Although DWC has not been studied for separation of benzene, EB, DEB,
PEB and inerts in the EB process, it has been studied for similar systems
Symbols
involving benzene and EB. Premkumar and Rangaiah (2009) reported
C Total production cost [US$]
significant energy and capital savings by using DWC. Gómez-Castro
Calloc Allocated cost of utility plants and related facil- et al. (2008) and Gutiérrez-Antonio and Briones-Ramírez (2009) opti-
ities [US$] mized DWC for multiple objectives. While EB process and DWC have
CB Equipment base cost [US$] thus been simulated and studied separately, complete EB process with
Ccatalyst Cost of catalyst [US$] DWC has not been investigated.
Ccont Cost of contingencies [US$] The present research has two key objectives. One is to develop and
Cd Discounted case flow [US$] simulate two complete processes for EB production, namely, liquid-
CDPI Direct permanent investment [US$] phase alkylation using zeolite catalyst, with CDCs or DWC employed
for downstream separation. Second objective is to analyse the two EB
Cexcl dep Total production cost excluding depreciation
processes for important objectives: total capital investment (CTCI ), net
[US$]
present value (NPV) and benzene loss, which encompass both eco-
Cland Cost of land [US$]
nomic and material indicators of the entire plant. Sensitivity of these
CP Purchase cost after inflation adjustment [US$] objectives with respect to lower/upper bound of decision variables is
CP,base Purchase cost at base year [US$] analysed by using an interface between Microsoft Excel and Aspen
Croyal Cost of royalties [US$] HYSYS. New contributions of the present study are development and
Cserv Cost of service facilities [US$] simulation of an improved, realistic and complete industrial EB process,
Csite Cost of site preparation [US$] considering reaction, separation and heat integration aspects. Selected
Cspares Cost of spare pumps [US$] alkylation and trans-alkylation reaction kinetics in the literature are
Cstartup Cost of plant startup [US$] validated with data from an industrial EB plant. Next, two different sep-
aration schemes, namely, CDCs and a DWC are studied for recovering
CTBM Total bare module investment [US$]
the product from the perspective of the complete EB process. Finally,
CTCI Total capital investment [US$]
heat integration and sensitivity analysis are conducted for both EB
CTDC Total depreciable capital [US$]
designs. This study shows that EB process using DWC instead of CDCs
CWC Working capital [US$] reduces CTCI by 2.8%, decreases benzene loss by 0.36% and increases
COM Cost of manufacture [US$] NPV by 138%. Hence, DWC is promising for improving sustainability of
D Depreciation [US$] the EB process.
f Fraction of total depreciable capital
GE General expenses [US$] 2. EB process design and simulation
FBM Bare module factor
NPV Net present value [US$]
An industrial EB process has 3 main stages: alkylation of ben-
r Interest rate [%]
zene to EB in several PFR beds, downstream separation of
S Total sales revenue [US$]
benzene, EB, DEB, PEB and inerts using two CDCs, and trans-
alkylation of DEB to EB in a single PFR bed. The process flow
diagrams of the complete processes employing either CDCs
have also introduced other alternative alkylation processes, but little or DWC are shown in Figs. 1 and 3, respectively. This study
information is known about them. Hence, liquid-phase zeolite-based uses plant capacity of 368,000 ton of EB/year, based on typi-
alkylation process is selected for the present study.
cal production rates of new EB plants in Asia Pacific (Woodle,
The EB process consists of reaction section (involving alkylation and
2006). Operating time is taken to be 8000 h/year and plant life
trans-alkylation reactions) and separation section to separate benzene,
is assumed to be 20 years. Minimum EB purity is 99.95 wt% for
EB, di-EB (DEB), poly-EB (PEB) and trace impurities such as methane
and ethane via distillation. Both Woodle (2006) and Welch et al. (2005)
the UOP EBOneTM process (Woodle, 2006). However, to satisfy
have summarized key design features of industrial EB plants. Luyben the contract liability in the event of disturbances, a slightly
(2002, 2010) has proposed two similar EB process designs, which use higher EB product purity of 99.97 wt% is set for the design.
continuous stirred tank reactors (CSTRs) instead of industrial practice Fresh benzene and polymer-grade ethylene are the raw
of using plug flow reactors (PFRs) and also ignore PEB formation (due materials for EB process. Fresh benzene is assumed to be
 144
chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157
Fig. 1 – Process flow diagram of EB production with liquid-phase zeolite-based reactions and conventional column separation; stream data (i.e., flow rate, temperature, pressure and
component mole fraction) and equipment specifications correspond to the average solution in Table 5.
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 145

Fig. 2 – DWC simulation using two columns in Aspen HYSYS.

at 25 ◦ C and 1 atm; its typical purity is 99.99 wt% (Folkins, relationship is using Ergun equation in Aspen HYSYS. DEB is
2005). Polymer-grade ethylene is 99.9 vol% of ethylene, with represented by 1–3 DEB as it is the most stable isomer and is
the remaining made up of ethane and methane in roughly present predominantly (Woodle, 2006). Other inputs required
equal proportions (Zimmermann and Walzl, 2007). Ethane and for reactor simulation in Fig. S1 are from Ganji et al. (2004).
methane are inerts in the EB process, thus necessitating a Results of reactor simulations using the three kinetic mod-
purge stream in downstream separation to prevent their accu- els are summarized in Table 2. Predictions of the alkylation
mulation. Zimmermann and Walzl (2007) state that ethylene reactors outlet temperatures by the three models are very
is typically transported in pipelines at pressures from 40 to close to one another and also to the plant data. Average per-
1000 bars. The lower limit of 40 bars is taken to avoid the criti- centage deviation is calculated ignoring the contributions of
cal pressure; since operation is below critical conditions, a feed temperatures (which are very close) and DEB weight fraction
temperature of 25 ◦ C is assumed to prevent liquid ethylene (which is small). Based on this and average absolute deviation
formation. for DEB, model 3 is better and hence selected for this study.
To determine an appropriate thermodynamic model for To design the alkylation and trans-alkylation reactors, various
simulations, validation is performed using vapour-liquid equi- other specifications are required as elaborated below.
librium (VLE) experimental data from Gmehling and Onken PFRs are chosen for alkylation and trans-alkylation as they
(2000). In this validation, only benzene/EB binary system is are industrially prevalent (Woodle, 2006). Subsequent analysis
considered due to unavailability of experimental data involv- performed in this study shows that the reactors tend towards
ing DEB. Both Soave–Redlich–Kwong (SRK) and Peng–Robinson smaller reactor diameter, allowing greater radial mixing and
(PR) equation of state models are considered as possible validating the PFR assumption. Adiabatic operation is selected
options based on Aspen HYSYS guidelines. Luyben (2002) sug- as recommended by Woodle (2006). Further, Woodle (2006)
gested Chao–Seader semi-empirical model, whereas Qi and recommends benzene to ethylene (B/E) ratio between 3 and
Zhang (2004) concluded that Benedict–Webb–Rubin–Starling 3.5 for liquid-phase zeolite-based alkylation. High B/E ratio
model can be utilized. Predicted data by these four models in improves benzene selectivity, reduces coke formation and
Aspen HYSYS are compared with experimental VLE data for helps in absorbing heat of reaction; however, it can also result
benzene-EB mixture at 1.013 bar. Root-mean-square errors for in lower benzene conversion and higher benzene recycle costs
X–Y and T–X data (see Table S1 in Supplementary informa- (Al-Kinany et al., 2012). The alkylation operating temperature
tion) show that both PR and SRK models give the best fit. PR is between 170 and 270 ◦ C (Woodle, 2006). High temperatures
model is finally selected as it can better predict liquid den- are necessary for increased catalyst activity, but this may lead
sities, especially for non-polar hydrocarbons, which form the to thermal runaways (due to the exothermic nature of alkyla-
main components in the EB process. tion) and to higher rates of side reactions, which deactivate the
catalyst, reduce EB selectivity and increase equipment cost.
2.1. Alkylation and trans-alkylation reactors Conversely, low temperatures reduce benzene conversion and
reaction rates, compromising on industrial feasibility. Woodle
Possible reaction schemes for alkylation and trans-alkylation (2006) suggests an operating pressure of 35–40 bars.
in the EB process are presented in Table 1. Rate equations Maintaining temperatures within operating limits in the
for Model 1 in this table are modified to give r1 , r2 and r3 alkylation reactor is crucial due to the exothermic nature of
in terms of kmol/m3 s. Further, models 2 and 3 differ in the the reaction. Three methods are employed for this (Welch
pre-exponential factor for r2 and r3 . Formation of PEB and et al., 2005; Woodle, 2006). Firstly, excess benzene is used to
by-products is neglected because reaction kinetics are unavail- absorb the heat of reaction, and unreacted benzene is recov-
able, liquid-phase zeolite-based reactions generally result in ered in downstream separation and recycled to the alkylation
decreased formation of by-products (due to lower operating reactor. Secondly, cold shot cooling is employed, in which
temperatures), and steric hindrance limits alkylation reaction ethylene feed is split across several alkylation reactor beds
rates inhibiting PEB formation (Woodle, 2006). arranged in series. Thirdly, inter-stage cooling is performed
The industrial reactors given in Ganji et al. (2004) are in heat exchangers. Woodle (2006) notes that two options are
simulated in Aspen HYSYS for validating kinetic models in typically practised in industries: one is a 4 reactor beds config-
Table 1. Alkylation is conducted in reactors R1(1) and R1(2) uration with 1 inter-stage heat exchanger, and another is a 6
while trans-alkylation is performed in reactor R2; each of these reactor beds configuration with 2 inter-stage heat exchang-
reactors has 2 or 3 beds as shown in Fig. S1 (see Supple- ers. A temperature sensitivity test is performed to analyse
mentary information). For these simulations, adiabatic PFRs the ability of both these configurations to maintain temper-
are employed as they are used in liquid-phase alkylation and atures. According to Woods (2007), temperature increase per
trans-alkylation (Woodle, 2006). Void fraction of zeolite cat- bed should not exceed 50 ◦ C for multi-bed adiabatic PFRs, to
alyst is 0.48 (Baerlocher et al., 2007). The pressure drop-flow minimize radial temperature gradients, and the reactor bed
 146
chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157
Fig. 3 – Process flow diagram of EB production with liquid-phase zeolite-based reactions and DWC separation; stream data (i.e., flow rate, temperature, pressure and component mole
fraction) and equipment specifications correspond to the average solution in Table 5.
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 147

Table 1 – Reaction schemes and rate equations adapted from the literature.
Reaction type Reaction rate (kmol/m3 s) Reactions

Alkylation r1 Benzene + Ethylene → EB

r2 EB + Ethylene → DEB
Trans-alkylation r3 DEB + Benzene  2EB

Reaction models Modified rate equations Source

 4

1.711×105 exp − 6.344×10 C
Model 1 r1 =  RT
3
E Ganji et al. (2004); catalyst densities
1−1.520×10−2 exp − 3.933×10
 4.703×104RT
CEB
(892.9 kg/m3 for alkylation and
r2 = 69.45 exp − 612.7 kg/m3 for trans-alkylation) are
 6.128×10RT4  CEB CE
40.47 exp − CDEB CB 1.0218 derived from the reactor data in this
RT
r3 = source.
1+3.000×10−6 CB
  4
Model 2 r1 = 1.528 × 106 exp − 7.113×10 C C Luyben (2002); activation energy is
RT
 4
 E B converted to J/mol.
r2 = 2.778 × 104 exp − 8.368×10 CEB CE
 6.276×104 RT
r3 = 0.4167 exp − RT CDEB CB
 4

Model 3 r1 = 1.528 × 106 exp − 7.113×10 C C Luyben (2010); activation energy is
 RT
4
 E B converted to J/mol.
r2 = 2.778 × 107 exp − 8.368×10 CEB CE
 6.276×104 RT
r3 = 1000 exp − RT CDEB CB

In this table, R is the molar gas constant in J/(mol K), T is the absolute temperature in K, and CE , CEB, CB and CDEB are concentrations of ethylene,
EB, benzene and DEB respectively in kmol/m3 . All rate equations in the cited papers are modified to give reaction rates in kmol/m3 s.

Table 2 – Comparison of simulation results by the three models with the plant data.
Plant Model 1 Model 2 Model 3

R1(1) Out ◦
T ( C) 252.31 252.00 252.10 252.00
Benzene (wt%) 89.17 89.75 89.71 90.11
EB (wt%) 9.86 10.17 10.29 9.21
DEB (wt%) 0.51 0.07 0.00 0.68
Ethylene (wt%) 0.00 0.00 0.00 0.00

R1(2) Out T (◦ C) 255.73 255.30 255.40 255.20

Benzene (wt%) 79.79 80.11 79.96 81.21
EB (wt%) 18.21 19.62 20.04 16.65
DEB (wt%) 1.33 0.27 0.00 2.15
Ethylene (wt%) 0.00 0.00 0.00 0.00

R2 Out T (◦ C) N.A. 217.90 217.90 218.20

Benzene (wt%) 93.69 95.25 95.43 93.71
EB (wt%) 3.46 0.87 0.39 5.06
DEB (wt%) 1.98 3.88 4.18 1.23
Ethylene (wt%) 0.00 0.00 0.00 0.00

Average % deviation (excluding T and DEB) – 14.74 17.64 10.71

Average absolute deviation for DEB – 1.13 1.35 0.58

outlet temperature should not exceed 270 ◦ C. Sensitivity test the industrial practice, a single bed is employed for trans-
results show that the 4 beds configuration has a bed temper- alkylation; this reactor is nearly isothermal as trans-alkylation
ature increase of close to 50 ◦ C for all beds, and to ensure has nearly zero heat of reaction.
that the outlet temperatures of the reactor beds are below For liquids over adiabatic fixed catalytic bed PFR, Woods
270 ◦ C, the inlet feed temperature range is small between (2007) suggested the ratios:
170 and 180 ◦ C. In contrast, the 6 beds configuration shows
greater leeway and flexibility, with a bed temperature increase H Dp
> 200 and < 0.10 (1)
of around 30 ◦ C and a larger feasible inlet temperature range Dp D
of 170–200 ◦ C. Since these facilitate process operation, 6 beds
configuration is selected, which is shown in Figs. 1 and 3.
Here, H = bed length, D = bed diameter and Dp = effective parti-
High benzene to DEB (B/DEB) ratio in trans-alkylation
cle size of the catalyst. Zeolite pellets of Dp = 5 mm are chosen
allows high DEB conversion to be achieved at equilibrium, but
for maximum exposure of the catalyst surface (Muthiya et al.,
at the cost of high recycle cost, whereas low ratio decreases
2014), Hence, H should be >1 m and D should be >0.05 m. Alky-
EB selectivity with the formation of heavier by-products.
lation and trans-alkylation reactor length and diameter are
Hence, Woodle (2006) suggests that the B/DEB ratio should be
then varied from these minimum values based on the fol-
between 2 and 4. Trans-alkylation reactor temperature range
lowing conditions. Reynolds number for fluid flow through
of 170–270 ◦ C is followed to maintain suitable catalytic activity
a packed bed should be more than 1000 for turbulent flow.
and reasonable reaction conversion, as well as ensure reac-
Ganji et al. (2004) noted that typical pressure drop in alkylation
tion is conducted in liquid phase (Woodle, 2006). Following
reactors is ∼200 kPa, and Luyben (2002) stated that ethylene
148 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157

pumps and vapour recompression in CDCs have been uti-

Table 3 – Specifications of component purities for
conventional distillations columns. lized successfully in petrochemical processes, and they have
potential for further reducing energy requirements in a CDC
Benzene column EB column
(Hanggi and Meszaros, 1999; Flegiel et al., 2015) and also in a
Distillate purity (%) EB: 0.05 (molar) EB: 99.97 (mass) DWC (Chew et al., 2014). Kiss (2011) studied heat integrated
Bottoms purity (%) Benzene: 0.01 (molar) EB: 0.05 (molar) RD for the synthesis of fatty acid methyl esters. These can be
analysed in the context of EB process, in future studies.
conversion per bed should be 100%. For the trans-alkylation
reactor, pressure drop should be below 300 kPa (Ganji et al.,
2004), and EB to DEB molar ratio in the outlet stream should 2.3. Separation using DWC
be around 3.5 (Luyben, 2002). With these considerations, alky-
lation reactor bed length and diameter are concluded to be A DWC is designed by combining the two conventional
7 m and 1 m respectively, while trans-alkylation reactor bed columns into one DWC, with EB produced in high purity in the
length and diameter are calculated to be 8 m and 2 m respec- side draw. It consists of a pre-fractionator and a main column.
tively; more details on this determination are given in Tables Feed to the DWC is the same as that to the benzene column.
S2 and S3 (see Supplementary information). EB product purity is specified as 99.97 wt%, both the distillate
and bottom streams of DWC as 0.05 mol% of EB and a vapour
2.2. Separation using CDCs distillate of flow rate 1.6 kmol/h. DWC has a partial condenser,
its operating pressure is the same as CDCs.
Separation is performed to separate benzene, EB, DEB and DWC is simulated in Aspen HYSYS using an equilibrium-
inerts (i.e., methane and ethane) and to recover EB product. For stage model. To obtain initial estimates for this, three shortcut
this, 2 distillation columns are considered as the formation of columns are constructed (Premkumar and Rangaiah, 2009).
PEB is neglected (Luyben, 2002). Direct sequence is employed The rigorous simulation consists of 2 distillation columns
based on heuristics in Seider et al. (2010). Feed to the first col- linked by material and energy streams (Fig. 2); it is known as
umn is from the alkylation and trans-alkylation reactor outlet acyclic system simulation. According to Navarro et al. (2012),
streams, after pressure is adjusted using liquid expanders for at the top of DWC, with the vapour stream “Overhead Vapour”
energy recovery. The first column is the benzene column to at saturation, added energy stream “e1” is equal to energy
separate excess benzene as the distillate. A purge stream is removed if a partial condenser is utilized in the main column
necessary to prevent accumulation of inerts in the system. of the DWC to provide reflux to the pre-fractionator. Likewise,
Since impurities are recommended to be removed early (Seider at the bottom of DWC, with the addition of saturated liquid
et al., 2010), a partial condenser is used on the first column. stream “Bottom Liquid” and energy stream “e3”, this energy
The vapour distillate having inerts and benzene is cooled and stream is the same as the energy added if a reboiler is uti-
sent to a flash drum. The liquid/benzene is recycled to the lized in the main column of DWC to provide vapour to the
alkylation reactor, while the vapour stream containing inerts pre-fractionator. Navarro et al. (2012) noted that all liquid and
and some benzene is used as fuel. The bottom stream from the vapour streams are in equilibrium, and their model (Fig. 2) is
benzene column is sent to EB column with a total condenser, helpful in ensuring convergence of simulation of DWC. Both
where EB product is obtained as the distillate. The bottom “Overhead Vapour” and “e1” enter at the same stage in the
stream of EB column is fed to the trans-alkylation reactor. main column. Similarly, both “Bottom Liquid” and “e3” enter
Based on guidelines/heuristics in Seider et al. (2010), oper- the main column at the same stage. A SET block is used to
ating pressure of benzene and EB columns is set at 30 psia equate energy stream “e3” to the negative of energy stream
(206.6 kPa), above the distillate bubble-point pressure. Con- “e2”, thus signifying that energy is transferred from the main
denser pressure drop is assumed to be 2 psia (14 kPa) while column to the pre-fractionator for providing the vapour split
column pressure drop is taken to be 10 psia (69 kPa), giving at the bottom of the DWC. Pre-fractionator column simula-
a reboiler pressure of 42 psia (289.6 kPa). The corresponding tion requires two degrees of freedoms, which are specified as
bubble-point temperatures of the bottoms are 185.4 ◦ C and condenser reflux ratio and reboiler vapour fraction, in order to
230.1 ◦ C for benzene and EB columns respectively, which are represent the vapour and liquid split in a DWC.
lower than the critical temperatures of EB and DEB of 344 ◦ C Preliminary analysis of CDCs and DWC was performed to
and 389.9 ◦ C (Daubert and Danner, 1989), thus validating the evaluate their relative feasibility and cost-effectiveness. This
feasibility of this operating pressure. is performed via one-variable-at-a-time approach to mini-
Trays are employed for all distillation columns as they mize total reboiler duty, which constitutes the bulk of total
are relatively cheap (Perry et al., 1997), used for atmospheric annual cost. DWC design is optimized by changing variables in
and above atmospheric conditions (Branan, 2005), suitable for accordance to those in Premkumar and Rangaiah (2009). Inter-
wider and taller columns, favoured at high liquid loads, more mediate liquid and vapour flows between pre-fractionator and
durable and easy to maintain (Kister, 1992), and generally con- main column of a DWC, are varied via reflux ratio and reboiler
sidered safe for maintenance. Sieve trays are selected due vapour fraction of the pre-fractionator column. For conven-
to their lower relative cost. Purity specifications for the two tional columns, only feed stage and total number of stages
columns are shown in Table 3. Flow rate of the vapour dis- are varied. The results in Tables S4 and S5 (see Supplemen-
tillate is initially set as 1.6 kmol/h based on mass balance of tary information) show that using DWC gives 31% savings of
inerts. Other initial estimates are obtained via the shortcut reboiler duty. A more detailed sensitivity analysis is performed
column using the Fenske–Underwod–Gillard method. Rigor- later for various important objectives. The process flow dia-
ous simulation is then performed via the equilibrium-stage gram of EB production using DWC, together with the design
model. The process flow diagram of EB production using con- and operating data corresponding to the average solution, is
ventional columns is shown in Fig. 1 along with design and shown in Fig. 3, which is different from Fig. 1 in the down-
operating data corresponding to the average solution. Heat stream separation section.
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 149

Isopescu et al. (2008) noted that DWC not only avoids Total bare module cost/investment (CTBM ) is given by:
the re-mixing effect encountered in CDCs, it also helps to
attain a high purity of the middle volatile product in a

CTBM = (CP FBM ) + Ccatalyst + Cspares (2)
three-component system. This is especially advantageous in
all equipment
the application of a Benzene-EB-DEB system, as EB can be
extracted at high purity, thus allowing industrial EB produc-
Here, CP = purchase cost after adjusting for inflation, and bare
tion to meet its purity requirement. In recent times, reactive
module factor FBM is given in Seider et al. (2010) except for
DWC has been implemented in a biodiesel process for fatty
liquid expander, which is taken to be 3.3. The last two terms
acid methyl esters synthesis, saving significant energy and
in the above equation are respectively for catalyst and spare
equipment costs (Kiss et al., 2012). Mueller and Kenig (2007)
pumps (one for each pump in the process). Zeolite catalyst
observed that substantial increase in reaction selectivity and
cost is taken to be US$25/kg (Silvy, 2004). Catalyst mass is cal-
separation efficiency can be achieved with the use of a reactive
culated from catalyst volume required, density and assuming
DWC.
void fraction of 0.48 (Baerlocher et al., 2007). It is assumed that
EBZ-500 (density = 550 kg/m3 ) is used for alkylation, while EBZ-
2.4. Overall design improvements 100 (density = 585 kg/m3 ) is used for trans-alkylation. With
catalyst life of 1.5 years (Bhatia, 1989) and plant operating life
Liquid expanders are used in place of valves to reduce pres- of 20 years, number of catalyst replacements is calculated to
sures of alkylation and trans-alkylation reactor outlet streams be 13. Thus, Ccatalyst can be calculated for the plant life of 20
to the inlet feed pressure for downstream separation, in order years. The other costs are calculated as follows.
to recover energy. They are added, as shown in Figs. 1 and 3,
satisfying the following criteria: inlet pressure is between CTDC = CTBM + Csite + Cserv + Calloc + Ccont (3)
100 and 3000 psig, pressure difference is between 100 and
2800 lbf /in2 , inlet flow rate is between 200 and 4000 gal/min, CTCI = CTDC + Cland + Croyal + Cstartup + CWC (4)
and energy recovered is between 50 and 3000 HP (Perry et al.,
1997).
Here, CTDC = total depreciable capital, CTCI = total capital invest-
Heat integration is performed for the entire plant to recover
ment. Cost of site preparation (Csite ) = cost of service facilities
and reuse heat, thus reducing hot and cold utilities required.
(Cserv ) = 5% of CTBM , cost of contingencies and contractor’s fee
As no process temperature exceeds 400 ◦ C, furnace is not
(Ccont ) = 35% of direct permanent investment (CDPI ). Cost of
required, only shell and tube heat exchangers are used, and
land (Cland ) = cost of royalties (Croyal ) = 2% of CTDC , and cost of
a minimum approach temperature of 10 ◦ C is applied, based
plant startup (Cstartup ) = 10% of CTDC . Allocated cost of utility
on heuristics. All process streams are first considered for heat
plants and related facilities (Calloc ) is found using correlations
integration, with the feasible streams concluded to be streams
in Seider et al. (2010).
before alkylation and trans-alkylation reactor beds. The tem-
Working capital (CWC ) is estimated by first calculating cost
perature differences and heat capacities of these four streams
of manufacturing (COM), total production cost (C) and total
are shown in Fig. S2 (see Supplementary information). Con-
sales revenue (S). COM is the sum of utility costs, labour
denser and reboiler duties are not considered in the heat
costs, maintenance costs, operating overhead, property taxes
integration due to potential control problems. Moreover, con-
and insurance, total depreciation via the straight-line method
denser and reboiler duties are significantly larger as compared
and total cost of feedstock. C is the summation of COM
to process stream duties; hence, most of the energy for con-
and general expenses (GE), assumed to be 11.55% of S. S
denser and reboilers has to be provided by external utilities.
is the sum of revenue from EB product sale, energy credit
With these considerations, a heat exchanger network shown
and fuel credit, all in US$/year. Energy credit is for energy
in Fig. S2 (see Supplementary information) is designed. In this
recovered by liquid expanders by taking electricity cost of
network, more than 90% of energy in hot streams is used for
US$0.060/kWh. Fuel credit for purge stream is calculated by
heating the cold streams, and no hot utility is required.
converting natural gas cost of US$0.136/SCM (Seider et al.,
2010) to US$0.01294/kWh using the lower heating value of nat-
ural gas (0.09514 SCM/kWh; Turton et al., 2013). Total heat
3. Equipment sizing and cost estimation
produced is calculated by factoring in the heat of combus-
tion of all individual components in the purge streams, thus
Sizing and cost estimation are primarily based on Seider giving the total fuel credit in US$/year. Unit costs of raw mate-
et al. (2010). An overdesign of 10% for each equipment size rials, EB and utilities are taken from Seider et al. (2010); see
is assumed as the safety margin (Towler and Sinnott, 2013). Table S8 in Supplementary information. Finally, CWC is the sum
The material selected for all process equipment is carbon steel of cash reserves (=COM/12), inventory (=[4/365] selling price
due to its compatibility with all chemical components present of EB + [2/365] cost of benzene), accounts receivable (=S/12)
and lowest cost. Sizing of all individual process equipment and accounts payable (=total cost of feedstock/12), and CTCI
are presented in the Supplementary Information. To adjust is determined via Eq. (4).
for inflation, current cost index, CEPCI is taken to be 600 com- To evaluate the profitability, net present value (NPV) is cal-
pared to 500 in base year 2006 given by Seider et al. (2010). culated by first determining net earnings and discounted cash
Equipment base cost (CB ) and base year purchase cost (CP,base ) flow (Cd ).
are calculated using equations in Seider et al. (2010); see Tables
S6 and S7 in the Supplementary information for details. Total  
Net Earnings = (1 − Tax Rate) S − Cexcl dep − D (5)
capital investment (CTCI ) is the one-time expense to design,
construct and start up an entirely new plant. This involves
different components as elaborated below. Cd = Net Earnings + D − fCTDC − CWC (6)
150 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157


t
Cd
Sensitivity analysis of the EB process is performed using
NPV = t
(7) an interface between MS Excel (via Visual Basic for Applica-
(1 + r)
t=0 tions) and Aspen HYSYS v8.4 (Sharma and Rangaiah, 2015)
by changing one decision variable at a time to its lower or
Tax rate is taken to be 17% while interest rate (r) to be 15%. Note upper bound while maintaining other variables at their nom-
that Cexcldep = total production cost excluding depreciation (D), inal values. This is mainly because optimal values of decision
f = 1/3 (assuming CTDC is split equally during the first 3 years of variables are often at their respective lower or upper bound.
construction), and t = number of years for construction + plant For sensitivity analysis, the initial population (of solutions or
operating life − 1. CWC is included at the end of the construc- individuals) is first generated at the upper and lower bounds of
tion period as a deficit, and returned as an asset at the end of the decision variables. These upper and lower bounds are suit-
the plant operating life. Plant startup cost and salvage value ably modified for simulation convergence and feasible ranges
are not considered. For the first 2 years of operation after plant of objective functions (e.g., positive NPV). Subsequently, one
startup, Cexcldep is taken to be 50% and 75% each year. Similarly, solution/individual at average values of both bounds is then
total sales (S) in the first 2 years are assumed to be 50% and 75% generated; these average values are the nominal values. Then,
each year. This is to account for gradual increase in produc- each DV is changed to its lower or upper bound and its
tion to the design capacity. Finally, NPV is the sum of present effect on objectives is studied. For this, the respective DV vec-
value of cash flow in each of t years (Eq. (7)). tors (solutions/individuals) are transferred from MS Excel to
Aspen HYSYS v8.4 for simulation before process variables are
extracted from the converged simulation and returned back
4. Sensitivity analysis
to MS Excel. The results are then used to calculate objective
functions and trend the effect of various DVs on the objective
The sensitivity analysis for EB process is performed by con- functions, in MS Excel.
sidering possible objectives and decision variables (DVs). By
analysing different objectives, meaningful conclusions can be
5. Results and discussion
obtained. As investment is limited and chemical plants are
profit-motivated, total capital investment (CTCI ) and NPV are
The two EB processes are analysed for 3 objectives: CTCI , NPV
selected as economic objectives. Material indicator, namely
and benzene loss. It is desirable to minimize CTCI , minimize
benzene loss in the purge stream, is also considered. These 3
benzene loss and maximise NPV for design of EB production
objectives are analysed by comparing their extent of change
plant. Sensitivity results are presented and discussed below.
with respect to changes in various DVs. This is performed
for both process designs (i.e., with conventional columns and 5.1. EB process with CDCs
DWC). As EB process is a large process, there are many pos-
sible DVs. The shortlisted DVs and their ranges are shown in Fig. 4 presents the effect of changing to lower/upper bound of
Table 4 for both EB processes. The feasible range of each DV is each of DVs on CTCI (Case 1), whose value at the average of
determined via operating limits and simulation convergence. bounds of all DVs is US$106.8 million, referred to as the nomi-
Sensitivity analysis is performed by considering all DVs and nal value. Alkylation reactor diameter (DV9) has large effect on
the 3 objectives. CTCI , perhaps due to use of 6 alkylation beds. CTCI at lower and
In Table 4, note that B = benzene, Split Ratio = ethylene feed upper bounds of this DV deviates from its nominal value by
split ratio, PF = pre-fractionator, Main Col = main column of −3.7% and 6.7%, respectively. Another DV having substantial
DWC and VentOut T = temperature of purge stream after cool- impact on CTCI is the number of stages in EB column (DV18);
ing. Feed Stage Ratio is used to find the feed stage as follows: CTCI at lower and upper bounds of this DV deviate from its
nominal value by −2.3% and 4.3%, respectively. This can be
Feed Stage = Round [Feed Stage Ratio × (Total Stages − 1) + 1] (8) attributed to the wide range taken for analysis, from 27 to 83
stages, which was intentional to reflect the range of stages in a
DWC, by adding the lower and upper bounds of N1, N2, N3 and
Here, Round implies rounding off to the nearest integer,
N4 as shown in Table 4. DV15 (B Column total stages) range is
and Total Stages represents number of trays (numbered from
chosen similar to DV18; however, range of stages for DV15 is
top to bottom) excluding condenser and reboiler. Total Stages
smaller, and so its effect on CTCI is narrower (Fig. 4).
and Feed Stage of a DWC are derived by dividing it into 4 dif-
Although CTCI at lower and upper bounds of each DV mostly
ferent sections of varying number of stages, as shown in Fig.
fluctuates about its nominal value, an anomaly for feed stage
S3 (see Supplementary Information). N1 denotes number of
ratio of the B column (DV16) can be seen in Fig. 4. Average feed
stages in the pre-fractionator, while N2 and N4 represent num-
stage ratio of 0.5 gives the lowest/optimal CTCI . Thus, it can
ber of stages in the top and bottom of the main column of the
be concluded that the optimal feed stage of B column should
DWC. As the middle section of the main column should be
not be too high or low but placed somewhere in the middle
comparable to that of the pre-fractionator, a variability of ±5
for good separation with lower reflux/boil-up. Optimal feed
trays, represented as N3, is assumed. This gives the number of
stage ratio of EB column for lower CTCI is near its upper bound.
stages in the middle section to be N1+N3. Hence, total num-
DV2 (inlet temperature to alkylation reactor bed 5), DV10 (first
ber of trays in the main column is N1+N2+N3+N4. The top feed
ethylene split ratio), DV17 (overhead vapour distillate) and
stage for the vapour stream from the PF to the main column is
DV23 (trans-alkylation reactor diameter) have lower CTCI at
N2, while the bottom feed stage for the liquid stream from the
their average values, and so their optimal values lie between
PF to the main column is N1+N2+N3. Lastly, side draw stage
their lower and upper bounds. DV22 and DV23 are for size
ratio is used to find the side draw stage in a DWC, as follows:
of trans-alkylation reactor bed. Larger trans-alkylation reactor
will result in a substantial increase in equipment cost, lead-
Side Draw = N2 + Round [Side Draw Stage Ratio × (N1 + N3)] (9) ing to higher CTCI and vice versa. However, a trans-alkylation
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 151

Table 4 – DVs and their ranges for EB process with CDCs and DWC.
EB process with CDCs EB process with DWC

Decision variable Lower Upper DV no. Decision variable Lower Upper DV no.
bound bound bound bound

Overall B/E ratio 3 3.5 1 Overall B/E ratio 3 3.5 1

R1B5 feed T (◦ C) 170 200 2 R1B5 feed T (◦ C) 170 200 2
R1B1 length (m) 1 5 3 R1B1 length (m) 1 5 3
R1B2 length (m) 1 5 4 R1B2 length (m) 1 5 4
R1B3 length (m) 1 5 5 R1B3 length (m) 1 5 5
R1B4 length (m) 1 5 6 R1B4 length (m) 1 5 6
R1B5 length (m) 1 5 7 R1B5 length (m) 1 5 7
R1B6 length (m) 1 5 8 R1B6 length (m) 1 5 8
R1 diameter (m) 1 2 9 R1 diameter (m) 1 2 9
Split ratio 1 0.15 0.25 10 Split ratio 1 0.15 0.25 10
Split ratio 2 0.15 0.2 11 Split ratio 2 0.15 0.2 11
Split ratio 3 0.15 0.2 12 Split ratio 3 0.15 0.2 12
Split ratio 4 0.13 0.17 13 Split ratio 4 0.13 0.17 13
Split ratio 5 0.13 0.17 14 Split ratio 5 0.13 0.17 14
B col total stages 24 42 15 PF total stages (N1) 17 49 15
B col feed stage ratio 0.3 0.7 16 PF feed stage ratio 0.2 0.8 16
B col vapour distillate (kmol/h) 0.7 2.5 17 PF reflux ratio 0.3 0.45 17
EB col total stages 27 83 18 PF vapour fraction 0.66 0.74 18
EB col feed stage ratio 0.3 0.7 19 Main col top section stages (N2) 8 12 19
B/DEB ratio 2 4 20 Difference in trays (N3) −5 5 20
R2 feed T (◦ C) 200 230 21 Main col bottom section stages (N4) 7 17 21
R2 length (m) 8 16 22 Main col side draw stage ratio 0.3 0.67 22
R2 diameter (m) 1.5 2.25 23 Main col vapour distillate (kmol/h) 0.7 2.5 23
VentOut T (◦ C) 40 45 24 B/DEB ratio 2 4 24
R2 feed T (◦ C) 200 230 25
R2 length (m) 8 16 26
R2 diameter (m) 1.5 2.25 27
VentOut T (◦ C) 40 45 28

Fig. 4 – Sensitivity analysis of CTCI with respect to 24 DVs for ethyl benzene process with conventional distillation columns.
See Table 4 for the significance of each DV on the x-axis.

reactor diameter that is too small may not be optimal, and a mum NPV simultaneously. Average of bounds of DVs gives NPV
balance has to be maintained for this reactor size. of US$3.62 million for the EB process with CDCs. NPV at lower
Fig. 5 shows NPV at lower bound, average and upper bound and upper bounds of alkylation reactor diameter (DV9) deviate
of each DV (Case 2). General trend of this figure is opposite to from its nominal value by 103.3% and −185.5% respectively, as
that of Fig. 4, confirming the inverse correlation of CTCI and a change in diameter has a significant effect on the alkylation
NPV; if CTCI is higher, NPV is lower and vice versa. Hence, each reaction conversion. The other significant DV affecting NPV
DV has an optimal value to obtain minimum CTCI and maxi- is DV16 (feed stage ratio of column B). This is different from
152 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157

Fig. 5 – Sensitivity analysis of NPV with respect to 24 DVs for ethyl benzene process with conventional distillation columns.
See Table 4 for the significance of each DV on the x-axis.

Case 1, in which DV18 (EB column total stages) is the most sig- to ethylene ratio for the alkylation reactor: B/E ratio) affect
nificant after DV9. While a greater number of stages leads to benzene loss by 2.2% and −1.3%, respectively. A lower B/E
increased column height, savings in vapourizing (in reboiler) ratio implies a lower benzene recycle, thus resulting in greater
and condensing (in condenser) can be achieved, thus resulting amount of benzene release in the purge stream. Similarly,
in a trade-off. NPV at lower and upper bounds of DV16 deviate lower bound of DV20 (inlet B/DEB ratio for the trans-alkylation
from its nominal value by −37.3% and −127.9%, respectively, reactor) increases benzene loss by 1.8%.
whereas 32.0% and −113.6% deviations are observed for lower
and upper bounds of DV18, respectively. Average of bounds for 5.2. EB process with DWC
DV16 is optimal, whereas the lower bound of DV18 leads to the
best outcome of low CTCI and high NPV. EB process using DWC for downstream separation is simi-
Similar to Case 1, there are some DVs (2, 10, 15, 17, 22 and 23) larly analysed for 3 objectives. Several derived variables are
having low NPV at their average of bounds. Other moderately also considered to reconcile differences between simulation
significant DVs include 15, 19, 20, 21, 22 and 23 (Fig. 5). As and actual DWC; reflux ratio and vapour fraction of the pre-
stated in Case 1 above, DV15, DV21, DV22 and DV23 affect CTCI fractionator (in Fig. 2) used in simulations are correlated to
significantly. Further, NPV at lower and upper bounds of DV20 liquid and vapour split to the pre-fractionator, respectively.
(inlet B/DEB ratio for trans-alkylation reactor) deviates from Aided by Fig. 7, their definitions are:
its nominal value by 17.3% and −32.6%, respectively. Higher
B/DEB ratio leads to higher recycle cost, resulting in lower NPV. Reflux Ratio = L/VC1
1 (10)
Fig. 6 presents the effect of lower bound, average and upper
bound of each DV on benzene loss (Case 3). Unlike Cases 1 and Liquid Split to Prefractionator = L/LC2 (11)
1
2, fewer DVs have significant impact on this objective. Average
 
values of all DVs give benzene loss of 6.10 kg/h for the EB pro- Vapour Fraction = V/ LC1
1 +V (12)
cess with CDCs. As expected, temperature of the purge stream
after cooling (DV24) affects benzene loss most significantly.
Vapour Split to Prefractionator = V/VC2
4 (13)
Benzene loss for lower and upper bounds of DV24 deviates
from its nominal value by −11.0% and 12.4%, respectively.
Fig. 8 presents effect of lower and upper bound of each DV
High value of DV24 promotes higher purge rate, thus leading
on CTCI , for EB process with DWC. Average values of DV bounds
to higher benzene loss, and vice versa. Lower bound of DV8
give CTCI of US$103.8 million (nominal value), which is lower
(length of the last alkylation bed) also has significant impact
than that for EB process with CDCs in Case 1. DVs present
on benzene loss by 8.4%. For lower value of DV8, alkylation
in both EB processes (i.e., with DWC and CDCs) exhibit simi-
reaction conversion will be lower, resulting in more benzene
lar trends. For DWC design, DV16 (pre-fractionator feed ratio)
leaving in the purge stream.
has the most impact on CTCI . For lower and upper bounds
Lower bound of DV17 (benzene column overhead vapour
of DV16, CTCI deviates from its nominal value by 0.12% and
distillate) decreases benzene loss by −5.3%; this is because of
7.9%, respectively. Similarly, DV9 (alkylation reactor diameter)
less benzene being carried over to the purge stream. However,
has significant effect on CTCI for DWC design. CTCI for lower
a higher vapour distillate can also result in lower benzene loss,
and upper bounds of DV9 deviates from its nominal value by
as the increased flow is sent to the alkylation reactor as ben-
−3.9% and 6.9%, respectively, similar to that for CDCs design.
zene recycle. Lower and upper bounds of DV1 (inlet benzene
Another DV with significant effect on CTCI is DV21 (number of
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 153

Fig. 6 – Sensitivity analysis of benzene loss with respect to 24 DVs for ethyl benzene process with conventional distillation
columns. See Table 4 for the significance of each DV on the x-axis.

Fig. 7 – Liquid and vapour flows between the pre-fractionator and the main column of DWC in Aspen HYSYS simulation.

stages at the bottom of the main column). CTCI for lower and uid and vapour splits of 0.412 and 0.608, respectively. Further,
upper bounds of DV21 deviates from its nominal value by 6.9% a negative value for N3 (DV20) gives a lower CTCI , implying
and 0.76%, respectively. DV15 (number of stages of the pre- that number of stages in the pre-fractionator has to be slightly
fractionator), which makes up a portion of middle section of more than that in the middle section of the main column.
the main column, is also significant in influencing CTCI (0.28% Fig. 9 depicts the effect of lower bound, average and upper
and 3.7% respectively for lower and upper bounds of DV15). bound of each DV on NPV, for DWC design. Similar to EB pro-
The average value of DV10 (first ethylene split ratio) gives cess with CDCs, NPV and CTCI correlate inversely. Average
lower CTCI compared to its lower/upper bound. This is simi- values of all DV bounds give NPV of US$8.60 million (nominal
lar to that in CDCs design, as alkylation reactor configuration value) for EB process with DWC. The most significant variable
remains the same. However, other 3 DVs (inlet temperature affecting NPV is DV16 (feed stage ratio of the pre-fractionator),
to alkylation reactor bed 5, overhead vapour distillate and which also has the most significant impact on CTCI . NPV values
trans-alkylation reactor diameter) do not have optimal CTCI for lower and upper bounds of DV16 deviate from its nomi-
at average values of their bounds, for DWC design as com- nal value by −2.6% and −137%, respectively. DV21 (number of
pared to CDCs design. In fact, both lower and upper bounds stages at the bottom of the main column) is also significant
of overhead vapour distillate have lower CTCI for DWC design in Fig. 9. NPV values for its lower and upper bounds deviate
compared to its nominal value. Liquid and vapour splits to the from nominal value by −129% and −6.8%, respectively. Hence,
pre-fractionator of DWC (calculated using Eqs. (11) and (13)) average values of DV16 and DV21 bounds give higher NPV.
do not have much effect, similar to reflux ratio (DV17) and Note that DWC design variables have significant effect on
vapour fraction (DV18) of the pre-fractionator. Average value both CTCI and NPV, whereas alkylation reactor diameter was
of reflux ratio and vapour fraction of 0.375 and 0.7 gives liq- the most significant for EB process with CDCs. DV15 (number
154 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157

Fig. 8 – Sensitivity analysis of CTCI with respect to 28 DVs for ethyl benzene process with DWC. See Table 4 for the
significance of each DV on the x-axis.

Fig. 9 – Sensitivity analysis of NPV with respect to 28 DVs for ethyl benzene process with DWC. See Table 4 for the
significance of each DV on the x-axis.

of stages of pre-fractionator, as well as middle portion of the bounds of vapour fraction (DV18) deviates from its nominal
main column N1), is also relatively significant in influencing value by −2.4% and 3.3%, respectively. Further, a negative N3
NPV for DWC design. NPV values at its lower and upper bounds (DV20) gives a higher NPV.
deviate from nominal value by −35% and −38%, respectively. Fig. 10 shows the effect of lower bound, average and upper
Hence, average value of DV15 bounds gives lower NPV. Average bound of each DV on benzene loss, for the EB process with
of bounds for DV27 (trans-alkylation reactor diameter) and for DWC. Average values of all DVs give benzene loss of 6.08 kg/h
DV28 (temperature of purge stream after cooling) gives lower for DWC design, which is slightly lower than that for EB pro-
NPV. The liquid and vapour splits to the pre-fractionator of cess with CDCs. Similar to EB process with CDCs, fewer DVs
DWC (related to reflux ratio – DV17 and vapour fraction – DV18, have significant impact on this objective. Generally, variables
respectively) have less effect on NPV, compared to other DVs that affect benzene loss in CDCs design have dominating effect
discussed. NPV value for lower and upper bounds of reflux for EB process with DWC also. DV28 (temperature of purge
ratio (DV17) deviate from its nominal value by 25% and −17%, stream after cooling) has the most significant impact on ben-
respectively. On the other hand, NPV value for lower and upper zene loss from EB process with DWC design. Benzene loss
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 155

Fig. 10 – Sensitivity analysis of benzene loss with respect to 28 DVs for ethyl benzene process with DWC. See Table 4 for the
significance of each DV on the x-axis.

Table 5 – Objective functions values for lower/upper bounds and average values of decision variables for EB process with
CDCs and DWC.
Objectives for lower and upper DV bounds Objectives for average value of DV bounds

CDCs DWC CDCs DWC

CTCI (Million US$) 102.8 to 114.0 99.8 to 112.0 106.8 103.8

NPV (Million US$) −3.10 to 7.36 −3.20 to 12.4 3.6 8.6
Benzene loss (kg/h) 5.43 to 6.86 5.41 to 6.83 6.10 6.08

for lower and upper bounds of DV28 deviates from its nom- objectives followed by a comparison of average solutions (i.e.,
inal value by −11% and 12%, respectively. This is expected as with all DVs at the mean of their respective lower and upper
temperature of purge stream directly affects overhead vapour bounds). Lastly, average solutions are further analysed in
distillate and benzene loss from EB process. Further, purge detail.
stream flow is negligible compared to other flows in EB pro- Table 5 compares ranges of objective functions for lower
cess, and so it does not have much impact on CTCI and NPV. and upper bounds on decision variables. CTCI is lower and
As for EB process with CDCs, lower bound of DV8 (length of NPV is higher for DWC design compared to CDCs design. This
the last alkylation bed) has significant effect on benzene loss, is generally due to use of one reboiler/condenser (instead of
affecting it by 8.6%. two reboilers/condensers in CDCs design), savings in reboiler
While benzene column vapour distillate flow rate signifi- and condenser duties, as well as the allocated cost of util-
cantly affects benzene loss of EB process with CDCs, its impact ities (Calloc ). This lowers utility/operating and capital costs
on EB process with DWC is less significant. This difference can substantially, which in turn, lowers COM, reduces total pro-
be attributed to different draw location of the purge stream. duction cost and increases NPV. Lastly, DWC design attains a
DV1 (inlet B/E ratio for the alkylation reactor section) is also slightly lower benzene loss, compared to CDCs design. Lower
relatively significant, as in CDCs design. Its lower and upper reboiler/condenser duties for DWC means lower vapour/liquid
bounds affect benzene loss from its nominal value by 1.8% loading, leading to decreases in distillate and purge stream
and −1.1%, respectively. Lower bound of B/DEB ratio for trans- flows.
alkylation reactor (DV24) is also notable, affecting benzene For the average values of DV bounds (Table 5), a 2.8% sav-
loss by 2%. As expected, liquid and vapour splits to the pre- ing in CTCI , 137.6% gain in NPV and 0.36% reduction in benzene
fractionator of DWC (DV17 and DV18), and N3 (DV20) are loss can be achieved by using DWC instead of CDCs. Reasons
relatively insignificant in affecting benzene loss, compared to for these are explained in the previous paragraph. Hence, EB
other DVs discussed. However, both lower and upper bounds of process with DWC is a better choice compared to that with
these DVs reduce benzene loss as compared to their averages, CDCs. In order to compare both EB process designs in detail,
albeit slightly, at around 0.035%. objectives for the average values of DV bounds are further
analysed. The main differences in the two designs are in the
5.3. Discussion number of columns, condensers and reboilers used; hence,
their bare module costs are compared in Table 6.
EB processes with CDCs and DWC are compared using the DWC design gives savings of ∼3% in bare module cost as
sensitivity analysis results, first for the ranges of various compared to CDCs design. As expected, it has a slightly larger
156 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157

Table 6 – Bare module cost comparison for columns and heat exchangers.
Equipment CDCs Equipment DWC

Column cost (in Benzene column 3921 DWC 8363

thousands of US$) EB column 4697
Total 8618 8363

Heat exchanger cost Benzene column condenser 173 Condenser 198

(in thousands of EB column condenser 79
US$) Benzene column reboiler 134 Reboiler 345
EB column reboiler 228
Total 614 543

Table 7 – Comparison of utility cost and allocated cost for utility facilities.
Cost Equipment CDCs Equipment DWC

Utility cost (thousands of Benzene column condenser 176 Condenser 217

US$/year) EB column condenser 75
Benzene column reboiler 2420 Reboiler 3479
EB column reboiler 1900
Total 4571 3696

Allocated cost (thousands of US$) – 11346 – 9786

diameter and much higher height compared to each of the Appendix A. Supplementary data
conventional columns. Diameter and height of DWC are 4.72 m
and 72.5 m respectively, while those of benzene column are Supplementary data associated with this arti-
4.11 m and 44.0 m respectively, and those of EB column are cle can be found, in the online version, at
3.20 m and 65.1 m respectively. However, DWC has 38% fewer [Link]
total number of trays compared to total number of trays for
benzene and EB columns together (namely, 127 after adjusting
for column efficiency). Thus, this leads to comparable cost of
columns. Further, DWC design involves 12% lower bare mod- References
ule cost for reboiler and condenser compared to that for CDCs
design. This is expected as the number of condensers and Al-Kinany, M., Al-Megren, H.A., Al-Ghilan, E.A., Edwards, P.P.,
reboilers required is halved. However, the largest contribu- Xiao, T., Al-Shammari, A.S., Al-Drees, S.A., 2012. Appl.
tion comes in allocated cost for utilities (Calloc ), which is given Petrochem. Res. 2 (3–4), 73–83.
Baerlocher, C., McCusker, L.B., Olson, D.H., 2007. Atlas of Zeolite
along with utility cost in Table 7.
Framework Types, 6th ed. Elsevier, Amsterdam.
As shown in Table 7, significant cost savings in terms of Bhatia, S., 1989. Zeolite Catalysts: Principles and Applications.
utility cost and Calloc are achieved by EB process with DWC; CRC Press.
total utility cost of condensers and reboilers is reduced by 19%, Branan, C., 2005. Rules of Thumb for Chemical Engineers.
and Calloc is reduced by 14%. These lead to significantly lower Elsevier, Amsterdam.
COM and total production cost (C) for the process. Since these Chew, J.M., Reddy, C.C.S., Rangaiah, G.P., 2014. Chem. Eng.
Process. 76, 45–59.
cost differences are over the plant life of 20 years, NPV change
Collins, D.E., Richey, F.A., 1992. In: Kent, J.A. (Ed.), Riegel’s
becomes very significant. Calloc (i.e., investment for building
Handbook of Industrial Chemistry. , 9th ed. Van Nostrand
offsite utility plants or upgrading related facilities) is directly Reinhold, New York.
correlated to the amount of utilities required. Hence, total Daubert, T.E., Danner, R.P., 1989. Physical and Thermodynamic
direct permanent investment (CDPI ) is reduced, resulting in a Properties of Pure Chemicals: Data Compilation. Taylor &
lower total capital investment (CTCI ) for the EB process with Francis, Washington D.C.
DWC compared to that with CDCs. Ebrahimi, N., Sharak, A.Z., Mousavi, S.A., Aghazadeh, F., Soltani,
A., 2011. Chem. Eng. Process. Process Intensif. 50 (1), 31–36.
Ebrahimi, N., Sharak, A.Z., Mousavi, S.A., Shahrak, A.Z.,
6. Conclusions Aghazadeh, F., 2012. J. Pet. Sci. Technol. 2 (1), 43–49.
Flegiel, F., Sharma, S., Rangaiah, G.P., 2015. Mater. Manuf. Process.
In this study, an EB production process is developed in detail 33, 444–457.
and systematically. Alkylation reactor design is evaluated rig- Folkins, H.O., 2005. Ullmann’s Industrial Chemistry, 7th ed.
Wiley-VCH.
orously for its kinetics and configuration; it is concluded that
Gómez-Castro, F.I., Segovia-Hernández, J.G., Hernández, S.,
kinetics reported in Luyben (2010) are better in simulating an Gutiérrez-Antonio, C., Briones-Ramírez, A., 2008. Chem. Eng.
industrial reactor, and that a system of 6-alkylation reactor Technol. 31 (9), 1246–1260.
beds with 2 inter-stage heat exchangers is optimal. Complete Ganji, H., Ahari, J.S., Farshi, A., Kakavand, M., 2004. Pet. Coal 46
separation system is designed using CDCs or DWC. Heat inte- (1), 55–63.
gration is finally performed. Sensitivity analysis of the three Gerzeliev, M., Khadzhiev, S.N., Sakharova, I.E., 2011. Pet. Chem. 51
(1), 40–49.
objectives: CTCI , NPV and benzene loss shows that CTCI is
Gmehling, J., Onken, U., 2000. Vapour–Liquid Equilibrium Data
reduced by 2.8%, benzene loss is lowered by 0.36% and NPV
Collection. DECHEMA, Frankfurt.
is increased by 137.6% by using DWC instead of CDCs in EB Gutiérrez-Antonio, C., Briones-Ramírez, A., 2009. Comput. Chem.
process, for average values of DV bounds. Hence, DWC has Eng. 33 (2), 454–464.
huge potential to improve sustainability of EB process. Hanggi, D., Meszaros, I., 1999. Sulzer Technical Review., pp. 32–34.
 chemical engineering research and design 1 1 8 ( 2 0 1 7 ) 142–157 157

Isopescu, R., Woinaroschy, A., Draghiciu, L., 2008. Rev. Chim. 59 Seider, W.D., Seader, J.D., Lewin, D.R., Widagdo, S., 2010. Product
(7), 812–815. and Process Design Principles: Synthesis, Analysis and
Kiss, A.A., 2011. Fuel Process. Technol. 92 (7), 1288-196. Evaluation. John Wiley & Sons, New Jersey.
Kiss, A.A., Segovia-Hernández, J.G., Bildea, C.S., Miranda-Galindo, Sharma, S., Rangaiah, G.P., 2015. In: Rangaiah, G.P. (Ed.), Chemical
E.Y., Hernández, S., 2012. Fuel 95, 352–359. Process Retrofitting and Revamping: Techniques and
Kister, H.Z., 1992. Distillation Design. McGraw-Hill, New York. Applications. John Wiley.
Luyben, W.L., 2002. Plant-wide Dynamic Simulators in Chemical Silvy, R.P., 2004. Oil Gas J. 102, 58–65.
Processing and Control. Marcel Dekker, New York. Towler, G.P., Sinnott, R.K., 2013. Chemical Engineering Design:
Luyben, W.L., 2010. AIChE J. 57 (3), 655–670. Principles, Practice and Economics of Plant and Process
Mueller, I., Kenig, E.Y., 2007. Ind. Eng. Chem. Res. 46 (11), Design, 2nd ed. Elsevier.
3709–3719. Turton, R., Bailie, R.C., Whiting, W.B., Shaeiwitz, J.A.,
Muthiya, S.J., Amarnath, V., Kumar, P.S., Kumar, S.M., 2014. Int. J. Bhattacharyya, D., 2013. Analysis, Synthesis, and Design of
Innov. Res. Sci. Eng. Technol. 3 (2), 33–37. Chemical Processes. Prentice Hall.
Navarro, M.A., Javaloyes, J., Caballero, J.A., Grossmann, I.E., 2012. Welch, V.A., Fallon, K.J., Gelbke, H.P., 2005. Ullmann’s Industrial
Comput. Chem. Eng. 36, 149–159. Chemistry, 7th ed. Wiley-VCH.
Netzer, D., Ghalayini, O.J., 2002. Hydrocarb. Process., 71–78. Woodle, G.B., 2006. In: Lee, Sunggyu (Ed.), Encyclopedia of
Odedairo, T., Al-Khattaf, S., 2010. Chem. Eng. J. 157 (1), 204–215. Chemical Processing, 1. Taylor and Francis Group, New York.
Perego, C., Ingallina, P., 2002. Catal. Today 73 (1–2), 3–22. Woods, D.R., 2007. Rules of Thumb in Engineering Practice.
Perry, R.H., Green, D.W., Maloney, J.O., 1997. Perry’s Chemical Wiley-VCH, Weinheim.
Engineers’ Handbook, 7th ed. McGraw-Hill, New York. You, H., Long, W., Pan, Y., 2006. Pet. Sci. Technol. 24 (9), 1079–1088.
Premkumar, R., Rangaiah, G.P., 2009. Chem. Eng. Res. Des. 87 (1), Zimmermann, H., Walzl, R., 2007. Ullmann’s Industrial
47–60. Chemistry, 7th ed. Wiley-VCH.
Qi, Z., Zhang, R., 2004. Ind. Eng. Chem. Res. 43 (15), 4105–4111.